Does the ICDS Improve Children’s Diets? Some Evidence from Rural Bihar

Abstract

Although there are several studies documenting the impact of the Integrated Child Development Scheme (ICDS) – the largest preschool intervention – in India, few have documented to what extent it improves the quantity and quality of food consumed by young children. This paper attempts to provide causal estimates of the impact of the ICDS on calories, protein, iron, and Vitamin A intakes of children. Using matching techniques to define an appropriate counterfactual, and a primary survey in four villages in rural Bihar, our results suggest that: (a) for older children three to six years who benefit from cooked meals, the ICDS did result in higher intakes of calories, protein, and iron, and no substantive evidence that as a consequence there was substitution away from food at home. However, there was no impact on vitamin A intake; (b) for younger children whose mothers are given take-home rations, there is no evidence the ICDS improved intakes of calories or any other nutrients. Thus, even though the monetary value of the transfer was the same across both age groups, there is evidence to suggest the mode of transfer does seem to matter to ICDS effectiveness, consistent with other literature.

Acknowledgements

Funding for the fieldwork was provided by ICRISAT. We are grateful to ICRISAT, NCAP, and ICAR-RCER, Patna and in particular to R.K.P. Singh, R. Padmaja, and Anjani Kumar for their support. Thanks are due to Deepti Goel, Anirban Kar, Uday Bhanu Sinha, and the referees of this journal whose comments greatly improved this paper. Last but not least, we are indebted to all the enumerators for painstaking data collection.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. World Health Organization. Nutrition: Micronutrient deficiencies. http://www.who.int/nutrition/topics/vad/en/, accessed on 14th June, 2016.

2. There are six components of the ICDS in all: supplementary nutrition, immunisation, health check-ups, growth monitoring, preschool education, and nutrition education to their mothers. This paper, however, focuses only on supplementary nutrition.

3. Other than anthropometric outcomes, studies involving the ICDS have also focused on education outcomes (Nandi & Laxminarayan, 2016) and on the role of anganwadi worker training in improving outcomes (Singh & Masters, 2016). Dreze and Khera (2017) have a review of the broader literature around the ICDS. For countries other than India, the literature on food intakes largely pertains to school-going children, and examines whether interventions succeed in transferring the entire amount of the intended transfer to the target population. This literature also makes a distinction between whether the transfers from targeted programmes are higher than what would be expected from an equivalent income transfer alone. For instance, while Jacoby, Cueto, and Pollitt (1996) find that the energy intake of children participating in a school breakfast programme in Peru increased by 50 per cent of the intended transfer, other studies find a higher impact. Studies by Ahmed (2004) for the United States, Murphy et al. (2003) for Kenya, and Jacoby (2002) for Philippines find that 50–100 per cent of the intended transfer sticks with the beneficiary child. However, results of Bhattacharya, Currie, and Haider (2006) are contrary to all other studies. Evaluating the school breakfast programme in the United States, they do not find any impact of participation on calorie intake of school going children. Adelman, Gilligan, and Lehrer (2007) conclude that the differences in the magnitude of impact across countries could be due to differences in the level of calorie intake before the programme was implemented. Food supplementation programmes have a higher potential of increasing the total calorie intake in countries which have a low level of food consumption.

4. The sampling strategy was to select from each sampled household one pre-school child; the further disaggregation into children above three years and those younger was done subsequently. Note, however, that (as expected) the distribution of households with children above three years of age, across size of land holding, is similar in both the census and sample. This is also true of households with children below three years of age.

5. A sample size of 100 children (spread over two groups) would have been powered to detect a difference of 250 calories, 6.8 grams of protein, and 1.9 mg of iron, but perhaps not of vitamin A, based on a first difference in means and assuming a 5 per cent probability of type I error and 20 per cent probability of type II error. These magnitudes represent half of the nutrient content of the ICDS meals, and was chosen to allow for the possibility of substitution from food provided at home.

6. Guidelines for Bihar are at: (http://www.icdsbih.gov.in/SupplementaryNutrition.aspx?GL=9&PL=8&SL=1).

7. The nutrient content of the ICDS meal can be calculated from the schedule of meals and weekly menu, which also specifies the quantity of each ingredient (per child) to be used in cooking. An average (per day per child) was then calculated over the week, weighted by the number of days each dish is served, results in the approximately 500 calories and 12–15 grams of protein that are supposed to be the norm.

8. At one of the centres, despite there being no appointed ICDS worker, SN services were provided with the help of other centres in the village.

9. We had to drop three observations for each of the sub-sample because we use characteristics of both parents to match SN participants with non-participants in our estimation and these children only had one parent alive.

10. Please refer to Mittal and Meenakshi (2016) for details of the dietary recall survey instrument.

11. We also carried out the analysis using the magnitude of intended transfer for adjusting the intake from ICDS meal for children aged three to six years. These results are not very different from the ones presented in the paper and will be made available on request.

12. Two of the dishes on the menu, khichdi and pulao, are supposed to include seasonal vegetables, the choice of which is at the discretion of the ICDS worker. We use an average of vitamin A and iron content of vegetables that were available during the time of survey.

13. As far as we are aware, the only way to accurately assess the energy/nutrient content of breast milk is to weigh the child before and after each feed; this was not feasible given our resources and field context.

14. Out of these 39 children, for 11 children, the day of the survey was Sunday and therefore ICDS meal was not available. We do not know the reason why other children did not consume the ICDS meal on the day of the survey.

15. The index for assets owned was constructed using information on ownership of assets such as farm implements, livestock, and consumer durables through Principal Component analysis (PCA).

16. The difference in the age of parents and mother’s rank (in terms of her status) in the household, both of which are exogenous, were used to create an index of mother’s bargaining power using PCA.

17. Several questions were asked to elicit mother’s nutritional knowledge. These included questions about awareness of vitamin A, iodine, and treatment of diarrhoea. All these variables were combined using PCA to create an index of nutritional knowledge.

18. Several other variables were examined, including: differences in the proportion of male children among surveyed children, birth order of the child, health endowment of the child, number of siblings, parents’ age, their literacy status, child care practices adopted, time spent in child care, presence of alternative caregiver, proportion of households with non-migrant father, household size, and land ownership. None of these variables was significantly different across participation status.

19. A comparison of the distribution of propensity scores also shows considerable overlap. All these results are available with the authors on request.

20. To assess if results are sensitive to the assumption regarding the energy content of a meal taken outside the home, all impact estimates were recomputed, assuming that the energy content of the meal was the lower bound of the 95 per cent confidence interval estimated by Tandon and Landes (2011). These results are largely unaffected (tables available with the authors on request).

21. The potential benefits that accrue to these children from other components of the ICDS, including vaccinations, health checks, and nutrition education (of their mothers) are not considered here.

22. A related literature has also examined whether the marginal propensity to consume from cash differs from in–kind transfers. For example Del Ninno and Dorosh (2003) find that for poor households, the marginal propensity to consume wheat from in-kind transfers was about 0.25, while that from cash income was near zero. Hidrobo, Hoddinott, Peterman, Margolies, and Moreira (2014) is another example that suggests that food transfers have a greater impact on calories, while cash vouchers impact diet quality.

Additional information

Funding

This work was supported by an ICRISAT VDSA grant.